Code

StreamHPC 2023-10-18#480

Code

Merged

stanleytsang-amd merged 107 commits intoROCm:developROCm/rocPRIM:developfrom StreamHPC:develop_stream_2023_10_18StreamHPC/rocPRIM:develop_stream_2023_10_18Copy head branch name to clipboardNov 14, 2023

+5,376-2,565

Commits

Commits on Oct 18, 2023

• Excessive shared memory usage in block_shuffle fix

  

  Melirius

  authored and

  Naraenda

  committed
  98994c5

• Merge branch '632-excessive-shared-memory-usage-in-block_shuffle-algo' into 'develop_stream'

  Show description for 0fbf43e
  

  Melirius

  authored and

  Naraenda

  committed
  0fbf43e

• remove block_sort_algorithm template param from block_sort_kernel_impl and block_sort_impl

  

  parbenc

  authored and

  Naraenda

  committed
  6c3d3c0

• fixed compile errors

  

  parbenc

  authored and

  Naraenda

  committed
  ab274c6

• Updated ChangeLog.md

  

  parbenc

  authored and

  Naraenda

  committed
  2329a5f

• remove unnecessary code

  

  parbenc

  authored and

  Naraenda

  committed
  69fbd4a

• fixed CHANGELOG.md to not be so verbose about non public api changes

  

  parbenc

  authored and

  Naraenda

  committed
  f68110f

• Merge branch '573-merge-sort-block-sort-remove-specializations-for-block_sort_algorithm-bitonic-and' into 'develop_stream'

  Show description for 6343c66
  

  parbenc

  authored and

  Naraenda

  committed
  6343c66

• Add dynamic dispatch and autotuning to device_adjacent_difference

  

  sbalint98

  authored and

  Naraenda

  committed
  22e731b

• Fix device_adjacent_difference storage type

  

  sbalint98

  authored and

  Naraenda

  committed
  18ede8e

• ci: remove autotune dependency from build:benchmark

  Show description for ad93f78
  

  Gergely Meszaros

  authored and

  Naraenda

  committed
  ad93f78

• Merge branch '607-tuning-adjdiff' into 'develop_stream'

  Show description for 2fa35ae
  

  Gergely Mészáros

  authored and

  Naraenda

  committed
  2fa35ae

• test: fix indexing error test_type_helper<custom_16aligned>::get_random_data

  Show description for 8783207
  

  Gergely Meszaros

  authored and

  Naraenda

  committed
  8783207

• Merge branch 'fixup_autotune' into 'develop_stream'

  Show description for ac746ed
  

  Gergely Mészáros

  authored and

  Naraenda

  committed
  ac746ed

• fixes for compilation in debug for radix_sort

  Show description for 343519e
  

  Gergely Meszaros

  authored and

  Naraenda

  committed
  343519e

• Merge branch 'fix_custom_16aligned_indexing' into 'develop_stream'

  Show description for 9cac25f
  

  Gergely Mészáros

  authored and

  Naraenda

  committed
  9cac25f

• fix: Detect DPP & DPP broadcast support with __GFX<GENERATION>__ macros

  Show description for 931280c
  

  Gergely Meszaros

  authored and

  Naraenda

  committed
  931280c

• refactor: Use __GFX<GENERATION>__ to detect NAVI cards

  

  Gergely Meszaros

  authored and

  Naraenda

  committed
  c46a941

• Merge branch 'debug_fixes' into 'develop_stream'

  Show description for 027d2ef
  

  Gergely Mészáros

  authored and

  Naraenda

  committed
  027d2ef

• docs: Update CHANGELOG for DPP & ROCPRIM_NAVI fixes

  

  Gergely Meszaros

  authored and

  Naraenda

  committed
  0a71e6e

• Merge branch 'dpp_broadcast' into 'develop_stream'

  Show description for 730664a
  

  Gergely Mészáros

  authored and

  Naraenda

  committed
  730664a

• remove deprecated structs and functions

  

  parbenc

  authored and

  Naraenda

  committed
  c7fea30

• rename scan_by_key_config_v2 to scan_by_key_config

  Show description for fc38c7b
  

  parbenc

  authored and

  Naraenda

  committed
  fc38c7b

• remove the option to use custom implemented config for histogram

  Show description for 668f5d3
  

  parbenc

  authored and

  Naraenda

  committed
  668f5d3

• update config compile time check to a different pattern

  

  parbenc

  authored and

  Naraenda

  committed
  c5195d9

• update documentation comments for configs

  

  parbenc

  authored and

  Naraenda

  committed
  ad58f6e

• change documentation comments

  

  parbenc

  authored and

  Naraenda

  committed
  1173bc5

• change documentation comments on device_radix_sort

  Show description for c28b2b2
  

  parbenc

  authored and

  Naraenda

  committed
  c28b2b2

• change documentation comment

  Show description for 97f48ee
  

  parbenc

  authored and

  Naraenda

  committed
  97f48ee

• update documentation comments

  Show description for 268a37c
  

  parbenc

  authored and

  Naraenda

  committed
  268a37c

• update documentation comments

  

  parbenc

  authored and

  Naraenda

  committed
  752aaea

• update documentation comments

  Show description for 31769d9
  

  parbenc

  authored and

  Naraenda

  committed
  31769d9

• remove wrap_adjacent_difference_config function

  Show description for 9711471
  

  parbenc

  authored and

  Naraenda

  committed
  9711471

• add missing transform_config ctor

  Show description for 733aec5
  

  parbenc

  authored and

  Naraenda

  committed
  733aec5

• fix binary search still using wrap_transform_config

  

  parbenc

  authored and

  Naraenda

  committed
  2aaaf87

• implement static_asset to make binary_search only use binary search configs, but also work with the underlying transform

  

  parbenc

  authored and

  Naraenda

  committed
  f753f37

• update changelog

  

  parbenc

  authored and

  Naraenda

  committed
  987f04e

• remove some *_v2s that went under the radar

  

  parbenc

  authored and

  Naraenda

  committed
  6c5ad41

• remove unnecessary default values

  

  parbenc

  authored and

  Naraenda

  committed
  103b892

• Merge branch '617-rocm-6-0' into 'develop_stream'

  Show description for 34c2ced
  

  parbenc

  authored and

  Naraenda

  committed
  34c2ced

• Add binary search, lower_bound and upper_bound documentation

  

  Beanavil

  authored and

  Naraenda

  committed
  9cb4ab2

• host_warp_size() is replaced with two different versions with parameters.

  Show description for be698b5
  

  parbenc

  authored and

  Naraenda

  committed
  be698b5

• comment out unused param names

  

  parbenc

  authored and

  Naraenda

  committed
  95a9991

• fix typos in the documentation

  

  parbenc

  authored and

  Naraenda

  committed
  ed788aa

• Merge branch '145-documentation-binary-search' into 'develop_stream'

  Show description for cfa1bdb
  

  Beanavil

  authored and

  Naraenda

  committed
  cfa1bdb

• move host_warp_size to config_type.hpp

  Show description for fe56926
  

  parbenc

  authored and

  Naraenda

  committed
  fe56926

• add error checks to host_warp_size calls in tests and benchmarks

  

  parbenc

  authored and

  Naraenda

  committed
  b524867

• fix format

  

  parbenc

  authored and

  Naraenda

  committed
  6063e8b

• add missing comment

  

  parbenc

  authored and

  Naraenda

  committed
  e902223

• fix error handling in lookback_scan_state.hpp

  

  parbenc

  authored and

  Naraenda

  committed
  3cc1b26

• fix compilation error

  

  parbenc

  authored and

  Naraenda

  committed
  b1df7a7

• Merge branch '624-remove-internal-references-to-host_warp_size' into 'develop_stream'

  Show description for 1b41ea4
  

  parbenc

  authored and

  Naraenda

  committed
  1b41ea4

• change block_radix_rank_match and block_histogram_atomic to use rocprim::match_any instead of implementing same functionality

  

  parbenc

  authored and

  Naraenda

  committed
  9707407

• change radix_digit_count_helper to use rocprim::match_any instead of implementing same functionality

  Show description for 7c33d35
  

  parbenc

  authored and

  Naraenda

  committed
  7c33d35

• add elect function to warp intrinsics

  Show description for b360815
  

  parbenc

  authored and

  Naraenda

  committed
  b360815

• update match_any to return 0 when predicate is false

  

  parbenc

  authored and

  Naraenda

  committed
  1c0dcaa

• fix the bit check in elect function

  

  parbenc

  authored and

  Naraenda

  committed
  5856531

• update changelog.md

  

  parbenc

  authored and

  Naraenda

  committed
  84f33e1

• fix hard coded warps per block value to come from param in kernel

  

  parbenc

  authored and

  Naraenda

  committed
  9c39575

• remove unused variables

  

  parbenc

  authored and

  Naraenda

  committed
  4f31872

• fix review comments

  Show description for b7c5363
  

  parbenc

  authored and

  Naraenda

  committed
  b7c5363

• update group_elect test

  Show description for 624777a
  

  parbenc

  authored and

  Naraenda

  committed
  624777a

• remove unnecessary comments

  

  parbenc

  authored and

  Naraenda

  committed
  eff7a21

• remove expected from group_elect test

  Show description for 4730dfc
  

  parbenc

  authored and

  Naraenda

  committed
  4730dfc

• fix overindexing

  

  parbenc

  authored and

  Naraenda

  committed
  a6e9d41

• fix review comments

  Show description for 64abd1c
  

  parbenc

  authored and

  Naraenda

  committed
  64abd1c

• format

  

  parbenc

  authored and

  Naraenda

  committed
  d62c807

• fix review comments

  

  parbenc

  authored and

  Naraenda

  committed
  2520342

• fix perf regression

  

  parbenc

  authored and

  Naraenda

  committed
  1c461d7

• undo group_elect in block_histogram_atomic.hpp, because of perf impact

  

  parbenc

  authored and

  Naraenda

  committed
  f6fc6b8

• fix bad func name in CHANGELOG.md

  

  Gergely Mészáros

  authored and

  Naraenda

  committed
  f1f8fe7

• fix merge errors

  

  parbenc

  authored and

  Naraenda

  committed
  ee63ca4

• Fix reduce_by_key algorithm so keys[0] is not flagged as a new run when is nan

  

  Beanavil

  authored and

  Naraenda

  committed
  be98ebf

• Merge branch '623-merge-the-many-copies-of-rocprim-match_any-code' into 'develop_stream'

  Show description for 46e0296
  

  parbenc

  authored and

  Naraenda

  committed
  46e0296

• Merge branch 'fix-RLE-for-nan' into 'develop_stream'

  Show description for 759a6f4
  

  Beanavil

  authored and

  Naraenda

  committed
  759a6f4

• make device_radix_sort compatible with compiler provided __int128_t and __uint128_t

  

  parbenc

  authored and

  Naraenda

  committed
  428f078

• add ifdefs to only compile int128 parts on clang/gcc

  

  parbenc

  authored and

  Naraenda

  committed
  8902238

• update changelog

  

  parbenc

  authored and

  Naraenda

  committed
  d793f43

• fix for int128 to_string labdas

  

  parbenc

  authored and

  Naraenda

  committed
  814d9c5

• add test for block_radix_sort int128 support

  

  parbenc

  authored and

  Naraenda

  committed
  8461645

• Implement block run length decode

  Naraenda

  committed
  7580bdd

• Merge branch '184-add-support-for-u-int128_t-in-deviceradixsort' into 'develop_stream'

  Show description for 38ba018
  

  parbenc

  authored and

  Naraenda

  committed
  38ba018

• Merge branch 'block-run-length-decode' into 'develop_stream'

  Show description for f7083e0
  Naraenda

  committed
  f7083e0

• Fix reduce_by_key algorithm so out of bounds items are not flagged as new runs for NaNs

  

  Beanavil

  authored and

  Naraenda

  committed
  53a3164

• Add reduce_by_key test to check that flagging is correct when keys are all different

  

  Beanavil

  authored and

  Naraenda

  committed
  43dc2f8

• Fix performance regression observed during tuning for gfx1030 and gfx1102

  

  sbalint98

  authored and

  Naraenda

  committed
  417cea1

• Merge branch 'fix-RLE-for-nan' into 'develop_stream'

  Show description for 8584dbf
  

  Beanavil

  authored and

  Naraenda

  committed
  8584dbf

• Block Runlength Decode: Fix incorrect offsets and improve test

  Naraenda

  committed
  b028d2e

• Merge branch '639-fix-regression-tuning' into 'develop_stream'

  Show description for e21a158
  

  sbalint98

  authored and

  Naraenda

  committed
  e21a158

• Remove duplicate key from .clang-format

  

  sbalint98

  authored and

  Naraenda

  committed
  e91e745

• Merge branch 'improve-block-rld-tests' into 'develop_stream'

  Show description for c39ee8c
  Naraenda

  committed
  c39ee8c

• Remove additional duplicates from clang-format

  

  sbalint98

  authored and

  Naraenda

  committed
  f6dda0a

• Merge branch 'fix-clang-format' into 'develop_stream'

  Show description for b3b8028
  

  sbalint98

  authored and

  Naraenda

  committed
  b3b8028

• Merge branch 'fix-clang-format-2' into 'develop_stream'

  Show description for a4c53f2
  

  matyas-streamhpc

  authored and

  Naraenda

  committed
  a4c53f2

• Fix binary_search upper/lower_bound config tuning

  Show description for 15adf4d
  

  sbalint98

  authored and

  Naraenda

  committed
  15adf4d

• unify language around config params in documentation

  

  parbenc

  authored and

  Naraenda

  committed
  04e4afb

• Merge branch 'fix-binary-search-tuning' into 'develop_stream'

  Show description for 15e1a67
  

  matyas-streamhpc

  authored and

  Naraenda

  committed
  15e1a67

• Merge branch 'fix_documentation' into 'develop_stream'

  Show description for 3a35351
  

  parbenc

  authored and

  Naraenda

  committed
  3a35351

• Make the autotune build job run nightly

  

  sbalint98

  authored and

  Naraenda

  committed
  7075051

• remove radix_sort_onesweep autotuning workaround

  

  parbenc

  authored and

  Naraenda

  committed
  56fe65d

• Merge branch 'add-tuning-build-to-pipeline' into 'develop_stream'

  Show description for 9487932
  

  Gergely Mészáros

  authored and

  Naraenda

  committed
  9487932

• Merge branch '626-check-skipped-device_radix_sort_onesweep-autotune-configuration-in-rocm-5-7-gfx906' into 'develop_stream'

  Show description for f760858
  

  parbenc

  authored and

  Naraenda

  committed
  f760858

Commits on Oct 20, 2023

• Resolve doxygen warnings for upstream PR

  Naraenda

  committed
  d07c302

• Merge branch 'fix-doxygen-warnings' into 'develop_stream_2023_10_18'

  Show description for 31f12a4
  Naraenda

  committed
  31f12a4

Commits on Nov 1, 2023

• Enable get_device_from_stream for Windows

  

  sbalint98

  authored and

  Naraenda

  committed
  8c66863

Commits on Nov 2, 2023

• Use _ENABLE_EXTENDED_ALIGNED_STORAGE for windows build in rmake.py

  Naraenda

  committed
  8c9c64c

• Bump unreleased ROCm version

  Naraenda

  committed
  466a946